The Changing Nature of Employment-Related Sexual Harassment: Evidence from the U.S. Federal Government, 1978 â•fi 1994

This paper examines the changing nature of attitudes toward and reports of sexual harassment using data for 1978–94 drawn from the U.S. Merit Systems Protection Board (USMSPB) of the U.S. federal government. The authors find that although unwanted sexual behavior reported by federal government employees changed only slightly in overall incidence over the period, its pattern changed noticeably. Unwanted sexual attention by supervisors, for example, declined in incidence; crude and offensive behavior by co-workers increased; and the likelihood that harassment would occur only once (rather than repeatedly) increased. Employees’ attitudes toward sexual harassment changed markedly, with a dramatically increased willingness to define unwanted sexual behavior as sexual harassment. This trend appears to have been due not to changes in employees’ demographic, human capital, and job characteristics, but rather to structural changes in their views of what constitutes sexual harassment. This article is available in ILRReview: http://digitalcommons.ilr.cornell.edu/ilrreview/vol57/iss3/7 443 Industrial and Labor Relations Review, Vol. 57, No. 3 (April 2004). © by Cornell University. 0019-7939/00/5703 $01.00 S THE CHANGING NATURE OF EMPLOYMENTRELATED SEXUAL HARASSMENT: EVIDENCE FROM THE U.S. FEDERAL GOVERNMENT, 1978–1994 HEATHER ANTECOL and DEBORAH COBB-CLARK* This paper examines the changing nature of attitudes toward and reports of sexual harassment using data for 1978–94 drawn from the U.S. Merit Systems Protection Board (USMSPB) of the U.S. federal government. The authors find that although unwanted sexual behavior reported by federal government employees changed only slightly in overall incidence over the period, its pattern changed noticeably. Unwanted sexual attention by supervisors, for example, declined in incidence; crude and offensive behavior by co-workers increased; and the likelihood that harassment would occur only once (rather than repeatedly) increased. Employees’ attitudes toward sexual harassment changed markedly, with a dramatically increased willingness to define unwanted sexual behavior as sexual harassment. This trend appears to have been due not to changes in employees’ demographic, human capital, and job characteristics, but rather to structural changes in their views of what constitutes sexual harassment. *Heather Antecol is an assistant professor of Economics at Claremont McKenna College, and Deborah Cobb-Clark is Director of the Social Policy Evaluation, Analysis and Research Centre at the Australian National University. A data appendix with additional results, and copies of the computer programs used to generate the results presented in the paper, are available from the first author at Claremont McKenna College, Dept. of Economics, 500 E. Ninth Street, Claremont, CA 91711. 1Fitzgerald and Shullman (1993) described sexual harassment as “a social problem with a long past and a short history.” exual harassment is a fact of life for many working women, with some studies suggesting that work-related sexual harassment may affect as many as one in two women at some point in their work lives (Schneider et al. 1997; Fitzgerald and Omerod 1993). The International Labour Organization (ILO), for example, recently reviewed the international literature and concluded that “sexual harassment is a pervasive problem affecting substantial numbers of women in every industrialized country where information is available” (ILO 1992). Sexual harassment is increasingly being recognized as an important economic issue in large part because of the substantial costs it imposes on workers and their employers. In spite of its pervasiveness, sexual harassment is not easy to define or measure. Many authors have suggested that the term “sexual harassment” is simply a new name for an old problem (ILO 1992; Fitzgerald and Shullman 1993),1 and despite the growing research, there is still no commonly accepted definition of sexual harassment 444 INDUSTRIAL AND LABOR RELATIONS REVIEW (Roscoe et al. 1994; Foulis and McCabe 1997). The existing evidence is often based on small, non-representative samples of women, hampering direct comparisons across studies (Fitzgerald and Shullman 1993). While some patterns in the factors associated with sexual harassment are beginning to emerge, we know almost nothing about how these patterns have changed over time as public awareness of sexual harassment has grown. The dearth of widely accepted stylized facts—along with a lack of agreement about how to define and measure sexual harassment—makes it difficult to address the problem and find solutions. This paper fills a void in the literature by using data for 1978–94 drawn from the U.S. Merit Systems Protection Board (USMSPB) of the U.S. federal government to examine the changing nature of sexual harassment. We focus on two dimensions of sexual harassment: the incidence of various unwanted sexual behaviors, and individuals’ views about what behaviors in fact constitute sexual harassment. Understanding workers’ perceptions of sexual behavior at work is especially important given a legal environment that relies on a reasonable victim standard—increasingly a reasonable woman standard—to make determinations in sexual harassment cases (Prior et al. 1997; Fitzgerald and Shullman 1993)2 and evidence that the negative consequences of unwanted sexual behavior at work are higher for women who believe themselves to be sexually harassed than for those who report unwanted sexual behavior but do not characterize it as harassment (Antecol and Cobb-Clark 2002). We are particularly interested in the following questions. To the extent that the incidence and definition of sexual harassment varied between 1978 and 1994, what drove this change? Was it a natural consequence of the altered demographic composition of the workplace that resulted from the expansion in employment opportunities for women? Or did it, rather, result from a change in the propensity—conditional on one’s characteristics—to experience unwanted sexual behaviors and label them as sexual harassment? Our goal is to answer these questions using standard decomposition and regression techniques that can identify the source of changes in sexual harassment over time. This—in combination with information about the qualitative nature of sexual harassment—sheds light on potential explanations for any structural shift in employment-related sexual harassment. Employment-Related Sexual Harassment: Evidence and Issues Sexual harassment cases first appeared in U.S. courts in the early 1970s when it was argued that sexual harassment constituted a form of gender-based discrimination. Since that time, public awareness of the issue has grown in large part due to certain well-publicized legal cases (Prior et al. 1997). To date, the study of employmentrelated sexual harassment has been mainly the purview of psychologists and sociologists. Economists have generally had relatively little to say about the matter. In particular, although several theoretical models of labor market discrimination exist in the economics literature, corresponding models of sexual harassment are notoriously absent.3 Most empirical research is based on surveys of selected workers, or, in some cases, university students. The small, non-representative nature of many of the samples used in these studies, as well as differences in survey design, methodology, and the way in which sexual harassment is measured, make synthesis of the results difficult. None2In particular, Equal Employment Opportunity Commission (EEOC) guidelines issued in 1980 emphasized that sexual harassment is unwelcome sexual behavior (emphasis added; Prior et al. 1997). 3The exception is Basu (2002), who modeled the circumstances under which it is Pareto improving to ban sexual harassment even though workers would find the pay attractive enough to submit to it. CHANGING NATURE OF SEXUAL HARASSMENT AT WORK 445 theless, several broad conclusions can be drawn from the existing literature. First, reports of sexual harassment are common across a number of employment situations in a number of countries. In particular, the U.S. evidence points to a high incidence of sexual harassment for women employed in a range of settings, including the military, large private-sector organizations, universities, legal practice, and the federal government (Antecol and Cobb-Clark 2002; Schneider et al. 1997; Laband and Lentz 1998; USMSPB 1995); and international research documents a high incidence of sexual harassment in many other countries around the world (Johnson 1994; ILO 1992). The incidence of sexual harassment is related both to demographic characteristics and to the nature of one’s employment. Women experience more sexual harassment than do men (see, for example, Fitzgerald and Ormerod 1993; Antecol and Cobb-Clark 2001), although many men also experience employment-related sexual harassment, and there is evidence that harassment of men is growing (USMSPB 1981, 1988, 1995). Victims of sexual harassment are also more likely than non-victims to be relatively young, to be unmarried, to have attended college, and to work exclusively with and be supervised by members of the opposite sex (see Fitzgerald et al. 1999b; Schneider et al. 1997; Laband and Lentz 1998; USMSPB 1995; Antecol and Cobb-Clark 2001). Organizational factors can also facilitate or inhibit sexual harassment (Williams et al. 1999). Women in the U.S. military, for example, report less sexual harassment when their duty stations have sexual harassment hotlines, offices devoted to investigating sexual harassment reports, and formal complaint procedures, or when they themselves have received sexual harassment training (Antecol and Cobb-Clark 2002). It is also important to note that while many workers say they have experienced unwanted sexual behavior, they often do not label their experiences as sexual harassment per se (see Antecol and CobbClark 2001; Marin and Guadagno 1999; Magley et al. 1999). Though the empirical evidence is limited, women appear more likely to apply the sexual harassment label when the behavior is more severe or when the gender mix in the work environment is not equal (Antecol and Cobb-Clark 2001). However, sexual harassment often goes unreported. Less than 5% of individuals experiencing sexual harassment ever report their experiences to anyone in authority, and even fewer file formal complaints with employers, institutions, or legal authorities (see Marin and Guadagno 1999 and Fitzgerald and Shullman 1993). Sexual harassment is particularly troubling in light of the mounting evidence that it has negative consequences for workers, including increased job turnover, higher absenteeism, reduced job satisfaction, lower productivity, and adverse health outcomes.4 Sexual harassment on the job also imposes sizable costs on firms. Between 1992 and 1994 sexual harassment is estimated to have cost the federal government $327 million (USMSPB 1995), and a study of 160 major U.S. firms found that sexual harassment costs each firm $6.7 million per year (not including the legal costs associated with defending such actions) (ILO 1992). The U.S. federal government makes an especially interesting setting in which to study employment-related sexual harassment. There has been a large expansion of female employment within the U.S. federal government—from 33% in 1978 to 44% in 1994 (U.S. Bureau of the Census 1980; USOPM 1999). Lewis (1996) reported that much of this expansion has occurred within traditionally male-dominated occupations—in particular, professional and administrative occupations—leading the gender integration of occupations to be much 4See Schneider et al. (1997), Fitzgerald et al. (1997), and Marin and Guadagno (1999) for reviews of the psychology literature regarding the consequences of sexual harassment. A small economics literature also examines the effect of sexual harassment on job satisfaction and intentions to quit (Laband and Lentz 1998; Antecol and Cobb-Clark 2002). 446 INDUSTRIAL AND LABOR RELATIONS REVIEW more rapid in the federal government than in the general economy. For example, in 1977, the average male federal government employee worked in an occupation that was approximately 78% male, whereas by 1993 this had fallen to 68%. The implications of these employment trends for sexual harassment are likely to be complicated. On the one hand, women have made rapid progress up the federal government’s occupational ladder, leaving many of them in high-level, supervisory positions and increasing their ability to influence institutional culture.5 At the same time, men and women are increasingly working together, which may increase the incidence of unwanted sexual behavior on the job. U.S. Merit Systems Protection Board Data We use data drawn from the U.S. Merit Systems Protection Board (USMSPB) of the U.S. federal government for 1978, 1987, and 1994. These data are uniquely suited to the analysis at hand. First, they provide us with fifteen years of comparable, consistently defined data on men’s and women’s experiences of and attitudes toward unwanted sexual behavior in the workplace.6 Additionally, whereas much of the existing employment-related sexual harassment literature is based on relatively small samples of workers in selected occupations (such as lawyers or academics), specific age groups (such as university students), or single firms, the USMSPB data set is large and encompasses public-sector workers employed in a range of occupations across all agencies of the federal government. In each of these years, a non-proportional, stratified sample of civilian employees was randomly drawn from the Central Personnel Data File (CPDF) operated by the U.S. Office of Personnel Management (OPM).7 Data were stratified on the basis of gender, agency, salary, and (in 1978 only) minority status. From an initial sample of 23,964 individuals in 1978, usable questionnaires were returned from 20,083 individuals, for an overall response rate of 84% (USMSPB 1981). In 1987 and 1994 the overall response rates were lower, with usable questionnaires returned from 8,523 and 8,081 individuals in those years respectively, for overall response rates of 66% and 61% (USMSPB 1988, 1995). Our final sample consists of 16,408 (1978), 7,487 (1987), and 5,875 (1994) civilian employees with non-missing values for the variables of interest.8 USMSPB respondents were asked whether they had experienced one or more of seven unwanted sexual behaviors on the job in the previous 24 months: (1) sexual gestures, (2) sexual remarks, (3) sexual materials, (4) pressure for sexual favors, (5) deliberate touching, (6) pressure for dates, and (7) sexual assault. Allowed responses include “never,” “once,” “once a month or less,” “two to four times a month,” and “once a week or more.” As our interest is in sexual harassment, we omit consideration of category (7).9 Our first step was to construct an indicator variable for each type of unwanted sexual behavior that 5Both supervisor’s gender and the gender composition of the work force are important determinants of the likelihood of being sexually harassed (USMSPB 1995; Fitzgerald et al. 1999b). 6We are aware of no other source of consistent data on sexual harassment spanning a similar time period. While cross-sectional analyses of USMSPB data exist (USMSPB 1981, 1988, 1995), these data have not been used to analyze the source of changes in sexual harassment over time. 7Some agencies—for example, the Central Intelligence Agency—are not required to report personnel information to the OPM. Therefore, civilian employees from these agencies are not included in the sample frame. For a list of excluded agencies, see USMSPB (1981, 1988, and 1995). 8Though the response rate is lower in the latter two years, the respondents were a representative crosssection of federal employees. Moreover, the latter two surveys were developed in part to examine the trends in sexual harassment in the federal government. For a more detailed discussion, see USMSPB (1988, 1995). 9Sexual assault involves rape or attempted rape and is not usually considered sexual harassment per se. CHANGING NATURE OF SEXUAL HARASSMENT AT WORK 447 equals 1 if the respondent reported experiencing the behavior at least once, and 0 otherwise. Second, we aggregated the responses to items (1)–(6) into three broad types of sexually harassing behavior: crude or offensive behavior (sexual gestures, sexual remarks, and sexual materials), unwanted sexual attention (pressure for sexual favors, deliberate touching, and pressure for dates), and any unwanted sexual behavior. Thus, following standard practice in this literature, our notion of sexual harassment is based on one or more experiences of unwanted sexual behavior. It does not rely on individuals reporting themselves to have been “sexually harassed” and does not necessarily fit with legal definitions.10 Respondents were also asked whether they would consider six separate unwanted sexual behaviors (specifically, sexual gestures, sexual remarks, sexual materials, pressure for sexual favors, deliberate touching, and pressure for dates) initiated by a supervisor to be sexual harassment. Identical questions were then asked about those same six behaviors initiated by a co-worker. Allowable responses include “definitely not,” “probably not,” “don’t know,” “probably yes,” and “definitely yes.” We constructed an indicator variable (Yit) for each of these twelve outcomes that equals 1 if respondent i reported in year t that he or she “probably” or “definitely” would consider that specific behavior to be sexual harassment and equals 0 otherwise. The reported incidence of unwanted sexual behavior and views about what constitutes sexual harassment are shown in Table 1. Not surprisingly, women were more likely than men to consider various unwanted sexual behaviors to be sexual harassment, regardless of the year.11 Further, the proportion of men and women who would view unwanted sexual behavior to be sexual harassment increased dramatically over time. For example, 91% of women in 1994 would have considered unwanted pressure for dates to be sexual harassment if initiated by a supervisor, compared to only 78% of women in 1978. Women were also more likely than men to report experiencing unwanted sexual behavior. In 1987, 41% of women reported experiencing any unwanted sexual behavior, compared with 15% of men. The most frequently reported form of unwanted sexual behavior was unwanted sexual remarks, while unwanted pressure for sexual favors was least common. Interestingly, the change over time in reported sexual harassment experiences is much smaller than the change over time in views about what constitutes sexual harassment. For example, the incidence of unwanted sexual gestures increased by 1.3 (1.2) percentage points for women (men) between 1978 and 1994, while the proportion of employees who would have considered unwanted sexual gestures from supervisors to be sexual harassment increased by 19.4 (18.2) percentage points for women (men) between 1978 and 1994.12 The Role of the Changing Composition of the Federal Work Force We can shed further light on these trends by assessing whether they can be explained by the changing composition of the federal work force. To this end, we analyze the twelve indicator variables (Yit) reflecting respondents’ attitudes about sexual harassment (see Table 1). The incidence of un10In particular, Fitzgerald et al. (1999a) argued that since without a judicial procedure it is impossible to determine who would meet legal criteria, a legal definition would be impractical for most research and policy purposes. Definitions based on the filing of a sexual harassment complaint are also flawed because sexual harassment often goes unreported, while definitions based on one’s perception of having been harassed introduce an element of subjectivity into the analysis (see also USMSPB 1995.) 11This is consistent with evidence that women see a wide range of behaviors as “harassing” (Fitzgerald and Shullman 1993), though the gender gap is generally smaller for more severe behaviors (Prior et al. 1997). 12USMSPB data also include demographic, human capital, and job information. See Appendix Table A1. 448 INDUSTRIAL AND LABOR RELATIONS REVIEW wanted sexual behaviors is captured by two of our broad measures of sexually harassing behavior: “any unwanted behavior” and “unwanted sexual attention” (see Table 1). This parameterization accounts to a degree for the severity of sexual harassment. The probability of viewing behavior j to be sexual harassment or experiencing unwanted sexual behavior is given by (1) Pr(Y j it = 1) = Pr(Xiβ + εit > 0) = Φ(Xiβ), where j indexes our 14 outcomes of interest and Φ is the standard normal cumulative density function. Equation (1) is estimated separately by gender and year using a probit model.13 The change between years t–1 and t in the expected probability of viewing a specific behavior as sexual harassment or in experiencing unwanted sexual behavior (Ŷ j) can then be approximated by (2) Ŷ j t – Ŷ j t–1 = ~ Φ(Xtβ̂t) – Φ(X – t–1β̂t–1) =~ [Φ(Xt β̂t) – Φ(X – t β̂t–1)] + [Φ(X – tβ̂t–1) – Φ(Xt–1β̂t–1)], Table 1. Reports of Attitudes toward and Incidence of Unwanted Sexual Behavior by Gender and Year. Proportion Reporting